Channel estimation enhanced LMS equalizer

Abstract

The present invention is related to an enhanced equalizer using channel estimation. A scaled version of a channel estimate is used as an expected average behavior of the product of a transmitted signal and a received signal to implement Griffith algorithm. The present invention also uses advance or prediction of a channel estimate to overcome the lag problem inherent in a least means square (LMS) algorithm in a time varying channel. Therefore, the present invention enables the use of a small step size while attaining the same tracking capability with a large step size. A channel estimate at some time in the future is used for updating equalizer filter tap coefficients. This may be performed with a prediction filter. Alternatively, a delay may be introduced in the input data to the filter tap coefficient generator, which makes a channel estimate look like a prediction to the filter tap coefficient generator.

1. A normalized least mean square (NLMS) equalizer comprising:
 a channel estimator configured to generate a channel estimate; 
 an equalizer filter configured to process samples of received signals with filter tap coefficients to generate filtered samples; 
 a delay buffer; and 
 a filter tap coefficients generator configured to update the filter tap coefficients based on the channel estimate, wherein the samples are delayed for a predetermined period of time before being forwarded to the equalizer filter and the channel estimate is processed for generating the filter tap coefficients without the delay for the predetermined period of time. 
 
   
   
     2. The equalizer of  claim 1  wherein the equalizer filter comprises:
 a tapped delay line, the samples being shifted into the tapped delay line sequentially; and 
 a vector inner product multiplier for computing a vector inner product of the samples shifted into the tapped delay line and the filter tap coefficients. 
 
   
   
     3. The equalizer of  claim 1  further comprising a down-sampler configured to down-sample the filtered samples, whereby the down-sampler outputs filtered samples at a chip rate. 
   
   
     4. The equalizer of  claim 3  further comprising a signal combiner for combining multiple streams of samples generated based on signals received through a plurality of antennas. 
   
   
     5. The equalizer of  claim 4  wherein the signal combiner interleaves the multiple streams of samples to generate a combined stream of samples. 
   
   
     6. A normalized least mean square (NLMS) equalizer comprising:
 a channel estimator configured to generate a channel estimate; 
 an equalizer filter configured to process samples of received signals with filter tap coefficients to generate filtered samples; 
 a delay buffer; and 
 a filter tap coefficients generator configured to update the filter tap coefficients based on the channel estimate, wherein the samples are delayed for a predetermined period of time before being forwarded to the equalizer filter and the channel estimate is processed for generating the filter tan coefficients without the delay for the predetermined period of time; 
 the equalizer filter comprises: 
 a tapped delay line, the samples being shifted into the tapped delay line sequentially; and 
 a vector inner product multiplier configured to compute a vector inner product of the samples shifted into the tapped delay line and the filter tap coefficients; 
 wherein the filter tap coefficients generator comprises: 
 a first conjugate unit configured to generate a conjugate of the channel estimate; 
 a second conjugate unit configured to generate a conjugate of a state vector of samples shifted into the tapped delay line; 
 a first multiplier configured to multiply descrambled filtered samples to the conjugate of the state vector; 
 an adder configured to subtract output of the first multiplier from the conjugate of the channel estimate to generate an unscaled correction term; 
 a vector norm square generator configured to compute a vector norm square of the state vector; 
 a divider configured to divide a step size parameter by the vector norm square to generate a scaling factor; 
 a second multiplier configured to multiply the scaling factor to the unscaled correction term to generate a scaled correction term; and 
 a loop filter for storing the filter tap coefficients of a previous iteration and outputting new filter tap coefficients by adding the scaled correction term to the filter tape coefficients of the previous iteration. 
 
   
   
     7. The equalizer of  claim 6  wherein the filter tap coefficients generator further comprises a third multiplier configured to multiply a leakage parameter to the filter tap coefficients of the previous iteration. 
   
   
     8. A method for performing equalization on received signals, the method comprising:
 generating samples of received signals; 
 generating a channel estimate based on the samples; 
 storing the samples in a delay buffer to delay the samples for a predetermined period of time before forwarding the samples to an equalizer filter; 
 processing the samples with the equalizer filter to generate filtered samples; and 
 generating new filter tap coefficients of the equalizer filter using the channel estimate, the channel estimate being processing without the delay. 
 
   
   
     9. The method of  claim 8  further comprising:
 down-sampling the filtered samples to generate filtered samples at a chip rate. 
 
   
   
     10. The method of  claim 8  further comprising:
 combining multiple streams of samples generated based on signals received through a plurality of antennas. 
 
   
   
     11. The method of  claim 10  wherein the multiple streams of samples are interleaved to generate a combined stream of samples.